Which path will NV40 use in Doom3?

[AlphaWolf]

So this means anyone with NV3x class hardware is SOL for Doom3, as we know they run the ARB2 path at half the speed of R3x0 class hardware. Looks like NV3x owners are being orphaned...

Clearly you don't know what you are talking about. Since day one, there has been no question that the NV3x flagship cards would run Doom 3 well compared to the competition. NVDA also has that Ultrashadow technology that is supposed to help speed up framerates in Doom 3. Heavy PS 2.0 usage is where the NV3x series falls short, but of course that doesn't really apply with Doom 3.

Carmack clearly stated r3xx would run arb path faster than nv3x. So what are you talking about?

<edit>although I don't think he ever said 1/2 of the speed

 

[Waltar]

I dunno if anyone bothered emailing him, but this is what John Carmack had to say:

[Doom 3 on the NV40 will use] ARB2. The NV40 is incredibly fast, far and away the fastest thing
you can
get right now.

I have removed the NV30 path, because Nvidia now has the ARB2 path
running
well on the earlier hardware.

John Carmack

Anyone else thinking Nvidia is using shader replacement to make the ARB2-path operate as if it was the NV30-path on earlier hardware?

If it's providing the same imagequality, does it matter? The ARB2 path was running at half the speed, a year and a half ago.. do you honestly think Carmack couldn't optimize his code for it? Do you honestly think nvidia drivers haven't gotten better with OpenGL?

 

[jimmyjames123]

Carmack clearly stated r3xx would run arb path faster than nv3x. So what are you talking about?

When I said "run well compared to the competition", I meant rull well with respect to framerates achieved. Since day one, NVDA was expected to do well with Doom 3 on their NV3x cards. For one, Doom 3 is not a heavy PS 2.0 type game, and also NVDA has the Ultrashadow technology that supposedly helps speed up framerates in shadow-intensive games like Doom 3. Initially, it was expected that the NV3x cards would use the NV3x path, and of course were expected to perform very well compared to the competition. Now, it seems that NVDA has found a way to boost performance on the ARB2 path (probably mainly due to improvements in the driver), and we all know that Carmack wouldn't remove the NV3x path if it resulted in the NV3x cards being trounced in Doom 3. So, IMO it is reasonable to suggest that the NV3x cards should perform well compared to the competition in Doom 3.

 

[digitalwanderer]

Ok I'll bite, could someone explain to me why they're running the ARB2 path well now? :|

*cough shader replacements cough*

?

Is that the only way? Couldn't they have finally figured out how to automagically optimize shaders at run-time? :|

 

[Kaotik]

So this means anyone with NV3x class hardware is SOL for Doom3, as we know they run the ARB2 path at half the speed of R3x0 class hardware. Looks like NV3x owners are being orphaned...

Clearly you don't know what you are talking about. Since day one, there has been no question that the NV3x flagship cards would run Doom 3 well compared to the competition. NVDA also has that Ultrashadow technology that is supposed to help speed up framerates in Doom 3. Heavy PS 2.0 usage is where the NV3x series falls short, but of course that doesn't really apply with Doom 3.

Carmack clearly stated r3xx would run arb path faster than nv3x. So what are you talking about?

<edit>although I don't think he ever said 1/2 of the speed

Actually he did, but that was over a year ago (Jan 29. 2003)

quote from his .plan:

The NV30 runs the ARB2 path MUCH slower than the NV30 path.
Half the speed at the moment. This is unfortunate, because when you do an
exact, apples-to-apples comparison using exactly the same API, the R300 looks
twice as fast, but when you use the vendor-specific paths, the NV30 wins.

 

[kipper67]

Do the first generation of the NV3X line (5200, 5600 and 5800) feature Ultrashadow ?

I thought it came in with 5900 and the models after that , 5700 and 5950.

Mark

 

[ChrisRay]

Do the first generation of the NV3X line (5200, 5600 and 5800) feature Ultrashadow ?

I thought it came in with 5900 and the models after that , 5700 and 5950.

Mark

No they do not.

 

[Xander]

Do the first generation of the NV3X line (5200, 5600 and 5800) feature Ultrashadow ?

I thought it came in with 5900 and the models after that , 5700 and 5950.

Mark

No they do not.

Really?

So, if the *nv30* - without Ultrashadow - is competitive in the ARB2 path...

Wow, the nv35 (WITH Ultrashadow) must simply pwn in it!

 

[Dave Baumann]

As I said ealier the "path" relates to the shader model for the unified lighting model - I don't see any reasons why UltraShadow is tied to the lighting shader and I would expect the engine just tests for the OpenGL extension anyway.

 

[Bouncing Zabaglione Bros.]

Clearly you don't know what you are talking about. Since day one, there has been no question that the NV3x flagship cards would run Doom 3 well compared to the competition. NVDA also has that Ultrashadow technology that is supposed to help speed up framerates in Doom 3. Heavy PS 2.0 usage is where the NV3x series falls short, but of course that doesn't really apply with Doom 3.

Clearly you don't know what you are talking about. You obviously never read Carmack's .plans where he tells us that on the standard Doom 3 path, NV30 was half the speed of R300. Carmack needed to code a special low-precision path for NV30 to get speed parity with R300.

 

[jimmyjames123]

Clearly you don't know what you are talking about. You obviously never read Carmack's .plans where he tells us that on the standard Doom 3 path, NV30 was half the speed of R300. Carmack needed to code a special low-precision path for NV30 to get speed parity with R300.

Actually, I do know what I'm talking about. In the beginning when Carmack made these comments, he mentioned that the NV30 was slightly faster on most scenes than the R300, using the NV3x path of course. At that time, the NV3x cards were never meant to be run on the ARB2 path because they gained a lot of performance using the NV3x path. However, this is the ill-fated NV30 that we are talking about, well before NVDA had their compiler technology anywhere near up to snuff, and NVDA at that time assured Carmack that there would be major improvements in performance once they started to further develop their compiler tech. Obviously, if the NV3x path really does get removed, then that shows that NVDA has made large strides in bumping up performance with ARB2, almost certainly largely due to driver improvements. And let's not forget that the NV35, as flawed as it is, was a major step up above the NV30 in part because of the 256 bit bus. Also seems that Ultrashadow technology was not fully developed when the NV30 was out, so that should enhance performance even further.

 

[Joe DeFuria]

Obviously, if the NV3x path really does get removed, then that shows that NVDA has made large strides in bumping up performance with ARB2, almost certainly largely due to driver improvements.

Yes and no.

Are they "improvements" or "hacks" that effectively translate the ARB2 calls the older NV3x calls?

In other words, is the ARB2 path going to render in flotating point precision all the time (as the R300, and presumably NV40) will? Or will the drivers "cheat" and throw in integer precsion, etc?

From a gamer perspective: I don't really care. Floating point precision isn't all that important for Doom3. (Not as important as being able to render in one pass per light, which the NV3x path allowed for anyway.)

But from a technology / analyis point of view, I have severe doubts that the NV3x rendering the ARB2 path would be apples to apples with the R3xx or NV4x in terms of the "work being done."

 

[jimmyjames123]

Tough to say, I guess. I am not even entirely confident that the NV3x path has been removed. Is there anyone who has been able to verify this?

 

[Bouncing Zabaglione Bros.]

Actually, I do know what I'm talking about. In the beginning when Carmack made these comments, he mentioned that the NV30 was slightly faster on most scenes than the R300, using the NV3x path of course. At that time, the NV3x cards were never meant to be run on the ARB2 path because they gained a lot of performance using the NV3x path. However, this is the ill-fated NV30 that we are talking about, well before NVDA had their compiler technology anywhere near up to snuff, and NVDA at that time assured Carmack that there would be major improvements in performance once they started to further develop their compiler tech.

You obviously don't know what you are talking about as it's been proved that the Nvidia "compiler tech" is as much hand tuned shader replacements inserted using app specific detection often using lower IQ, as actual legitmate optimisations.

If NV30 has suddenly doubled it's speed after a year, you can guarentee it's not just because of true optimisations in the drivers. The limitations of the hardware have not changed.

 

[jimmyjames123]

You obviously don't know what you are talking about as it's been proved that the Nvidia "compiler tech" is as much hand tuned shader replacements inserted using app specific detection often using lower IQ, as actual legitmate optimisations.

You are obviously so blinded by your ATI fanboyism that you can't use simple logic to properly conduct an argument regarding NVDA. The NVDA compiler works, has been tested by numerous reviews, and does not degrade iq as far as I know. If you could point to some conclusive evidence that the compiler degrades iq, then I would like to see it.

Your initial argument was: "So this means anyone with NV3x class hardware is SOL for Doom3, as we know they run the ARB2 path at half the speed of R3x0 class hardware". Obviously that is an ignorant statement, for the reasons I gave you. You are still stuck in 2002/2003 unfortunately. Open your eyes already.

If NV30 has suddenly doubled it's speed after a year, you can guarentee it's not just because of true optimisations in the drivers. The limitations of the hardware have not changed.

The NV30 was replaced by the NV35, which had some superior hardware attributes vs the NV30, and also has implementation of the Ultrashadow technology. You are not really making a good argument as to why the current NV3x cards would do poorly in Doom 3. Face the facts, we are not talking about a PS 2.0 shader heavy game here.

 

[Bouncing Zabaglione Bros.]

You are obviously so blinded by your ATI fanboyism that you can't use simple logic to properly conduct an argument regarding NVDA. The NVDA compiler works, has been tested by numerous reviews, and does not degrade iq as far as I know. If you could point to some conclusive evidence that the compiler degrades iq, then I would like to see it.

Search the forum. Was done to death when Nvidia claimed that the patch for 3DMark2003 that changed the shaders to prevent app-specific detection broke their "compiler". Nvidia were doing app specific detection and swapping out shaders for hand tuned low precison ones.

Your initial argument was: "So this means anyone with NV3x class hardware is SOL for Doom3, as we know they run the ARB2 path at half the speed of R3x0 class hardware". Obviously that is an ignorant statement, for the reasons I gave you. You are still stuck in 2002/2003 unfortunately. Open your eyes already.

Because I was referring to NV30 vs R300. Those with 9700Pros will still be able to run D3. Those with 5600's only get half the speed on the ARB2 path. That's okay because Carmack had a special NV30 path, but now that gone. So where has this improvement come from without a PP path in the game? PP in the drivers? Is it even the ARB2 path anymore if the drivers are doing code replacement to PP behind the scenes?

The NV30 was replaced by the NV35, which had some superior hardware attributes vs the NV30, and also has implementation of the Ultrashadow technology. You are not really making a good argument as to why the current NV3x cards would do poorly in Doom 3. Face the facts, we are not talking about a PS 2.0 shader heavy game here.

And NV35 is being replaced by NV40 - what's your point? I was talking about NV30, not NV35. Shaders are irrelevent - it's Carmack who says NV30 is half the speed of R300 using the ARB2 path. It's the people with NV30 and to a lesser extent NV35 that will suffer poor performance if they have to run ARB2 instead of a special NV3x path. It's these people that are being orphaned if there is no support in the game.

And yet Nvidia manages to suddenly double the speed in ARB2 after more than a year of driver development - that doubling of speed hasn't happened in the hardware.

 

[Bjorn]

Because I was referring to NV30 vs R300. Those with 9700Pros will still be able to run D3. Those with 5600's only get half the speed on the ARB2 path. That's okay because Carmack had a special NV30 path, but now that gone. So where has this improvement come from without a PP path in the game? PP in the drivers? Is it even the ARB2 path anymore if the drivers are doing code replacement to PP behind the scenes?

I'd hoped that Carmack would have commented on this when he was saying that the ARB2 path was running at full speed now. That he didn't say anything might of course be because it runs in full quality (which i definitely think is possible since it's not a shader heavy game afaik) but it would be good if anyone confirmed this. Although we will of course be able to verify that when the game is released.

 

[jimmyjames123]

Search the forum. Was done to death when Nvidia claimed that the patch for 3DMark2003 that changed the shaders to prevent app-specific detection broke their "compiler".

We are not talking about 3dmark03 here, we are talking about an actual game that is not a PS 2.0 shader heavy game! Get with/in the game already!

Because I was referring to NV30 vs R300. Those with 9700Pros will still be able to run D3. Those with 5600's only get half the speed on the ARB2 path. That's okay because Carmack had a special NV30 path, but now that gone. So where has this improvement come from without a PP path in the game? PP in the drivers? Is it even the ARB2 path anymore if the drivers are doing code replacement to PP behind the scenes?

You said NV3x vs R3xx. You have to wake up and realize that half speed of the NV30 when using ARB2 was noted a long long time ago, possibly more than a year ago. The 5600, 5800, and 9700Pro are hardly even benchmarked anymore. You should be talking about the 5700, 5900, and 9800Pro.

And NV35 is being replaced by NV40 - what's your point?

LOL, are you serious? We are talking about the NV3x series here. The NV30 had a very limited life span, for good reason, and very very few people even bothered to purchase one.

And yet Nvidia manages to suddenly double the speed in ARB2 after more than a year of driver development - that doubling of speed hasn't happened in the hardware.

Who made the claim that NVDA has managed to double their speed in ARB2? It would be naive to think that the performance has not jumped way up on Doom 3 from NV30 to NV35 though, mainly because of the 256 bit bus and Ultrashadow technology.

You should wait for some Doom 3 data and iq analysis on the NV3x cards before jumping to such negative conclusions. The architecture is very well suited for a game like Doom 3.

 

[Bouncing Zabaglione Bros.]

LOL, are you serious? We are talking about the NV3x series here. The NV30 had a very limited life span, for good reason, and very very few people even bothered to purchase one.

Well duh, I am talking about those people with poor performing cards that needed their own special path. I'm talking abou those people that have now been orphaned if the NV3x special path has now been removed from the Nvidia killer app. To talk about the NV40 being better is irrelevent - which is why I'm not talking about it.

I guess your attitude matches that of the company you love. First let's pretend the NV30 never existed, then let's ignore all the people who bought NV3x and stop bothering to support them with special paths. Then tell them they should forget about the money they spent on NV3x cards and tell them to buy NV40 instead.

And you have the gall to call me a fan boy

 

[Zeross]

This is becoming really difficult to understand :?
first it was "Doom III performance can't be compared between NV30 and R300 because it's not the same code path used" now that NV3x are using ARB2 and Carmack himself said it was fast enough even on older chips there is still something wrong ?

I don't get it ? what's wrong ? Don't you think that if Carmack dropped NV30 code path which was already written and good performing it's for a good reason ? I mean you don't throw your work away like that except if it is now useless !

guess your attitude matches that of the company you love. First let's pretend the NV30 never existed, then let's ignore all the people who bought NV3x and stop bothering to support them with special paths

Please explain me I'm lost : it's Carmack who is coding Doom III right ? Nvidia has nothing to do with his code right ? If he decides to drop a code path the decision comes from him and has nothing to do with Nvidia so I can't see how you could blame Nvidia for that !